This week, Facebook announced that it is introducing ‘automatic alternative text.’
This amazing new development is able to generate a description of a photograph on the social network, allowing visually impaired users to get an idea of what each picture contains.
Previously, people using screen readers would hear the name of the original poster followed by the term ‘photo’ – now they will hear ‘image may contain three people, smiling, outdoors.’
Using Facebook’s object recognition technology, the software is based on a neutral network with billions of parameters and trained with a million of examples.
And we think that’s pretty amazing!